In a high-temperature membrane reactor, one of the reaction products is selectively removed from the reaction mixture, thus preventing the mixture from reaching equilibrium. In a previous study [I], a CVI-silica membrane was used for the direct dehydrogenation of propane in a high-temperature catalytic membrane reactor This H-2 selective membrane had only a moderate permeation (similar to 140 x 10(-9) mol/m(2)Pa s) and a limited H-2/C3H8 permselectivity (alpha(0) approximate to 70-90 at 500 degrees C). These experiments proved that (at 500 degrees C) the propane conversion could be improved from the equilibrium value (similar to 18%) to a value which is about twice as high. The increase was however only significant for relatively small values of the propane feed stream less than or equal to 16.5 mu mol/s. This is because at high propane feed, the hydrogen cannot be removed fast enough through the membrane and conversion is again limited by the thermodynamic equilibrium. In this study, the comparison is made between the performance of the CVI-silica membrane and a Pd/Ag membrane when used as the H-2 selective membrane. The performance of the Pd/Ag membrane is far superior to the performance of the SiO2 membrane. H-2 fluxes of more than 0.1 mol/m(2)s were measured and the H-2/Ar permselectivity exceeds 4500. When it is run under comparable conditions, the performance of the Pd/Ag membrane reactor is much better. The increase in propane conversion persists at values of the propane feed stream that are about six times higher (105 mu mol/s). Since the: H-2 is selectively removed from the reaction mixture, it is not available for any competitive side reactions. The production of methane, which limits the propene selectivity of the reaction in a conventional plug-flow reactor, is much less in a catalytic membrane reactor. This means that the selectivity in the membrane reactor is higher than in the plug-flow reactor when they are fun under similar conditions. (C) 2000 Elsevier Science B.V. All rights reserved.